1

Authors: H. Rashvand, C. Khirallah, V. Stankovi, and L. Stankovi

Chapter 12:
Cooperative Network Coding
Cooperative Networks
Editors: M. S. Obaidat and S. Misra

Publisher Wiley
2
Outline
Introduction
Network coding (NC) concept
Cooperative relaying
Cooperation strategies
Performance measure
High SNR regime
Low SNR regime
Cooperation via Network Coding (CNC)
Introduction
Most dynamic and unstructured networks with distributed sources
and destinations are wireless and due to distributed variable
interference conditions they suffer from heavy outage and
extensive loss of data.
Diversity, which is an inherent part of the broadcasting nature of
wireless channels, is often used to combat fading effects by
increasing the signal-to-noise ratio (SNR) of the received signals.
Diversity is achieved using several schemes such as frequency,
time, polarization, space, multi-user, cooperative diversity.
Diversity comes at the cost of additional complexity to the network
routing process and effective node capacity as well as increasing
waste of bandwidth resources due to additional overheads, and the
need for multiple antennas for spatial diversity (e.g., multiple-input
multiple-output MIMO).
3
Introduction
Cooperative communications networks improve the system
performance in terms of:
reduced power consumption, increased system capacity and greater
resilience.
For example, physical layer cooperation among wireless nodes or
users resemble virtual MIMO Spatial diversity gain or Rate
multiplexing gains.
4
Network coding (NC)
NC is proposed by Ahlswede [1] as an generalised evolution of the
simple network routing protocols.
NC allows nodes in the network to go beyond the simple
forwarding of received messages to further mix (or encode) these
messages, at the intermediate nodes, before forwarding them to
destination, or decode these messages at destination nodes,
NC differs from simple network routing due to the fact that NC
aims at whole network optimization rather than individual classes
of users or applications being the objective,
Unlike traditional use of error control coding in the networks
where coding is performed at the edges (end-to-end) to detect
and/or correct the errors on individual packets for a given link, NC
seeks to combine different diversified routes in a multi-path
network routing fashion at the network level for the purpose of
better usage of network resources.
5
Network coding (NC)
6
Figure 12.1 shows a NC example where:
Two sources nodes X1 and X2 want to communicate two messages A and B
with out-of-range destination nodes Y1 and Y2, via nodes J and K,
Sending A and B directly (without NC) will require the use of 4 separate
links, X1-Y1, X1-Y2, X2-Y1 and X2-Y2 while,
Sending A and B using NC will require one shared link J-K, to transfer NC
encoded A+B, and two side links X1-Y1 and X2-Y2,
NC for this 2-by-2 nodes setup 1/4 = 25% save in links.

Figure 12.1. NC on
a 2-by-2 butterfly network[1]
(IEEE copyright line 2010).

Cooperative relaying
The information-theoretic properties of classic three-node relay
network, shown in Figure 12.3, with a source (S) that transmits
information, a destination (D) that receives information and a relay
(R) that both receives and transmits can be traced back to the
seminal work of Cover and El Gamal [7]. This work analyses the
relaying capacity under an additive white Gaussian noise (AWGN)
relayed channel, and comes up with several optimum relaying
strategies.
7
In Figure 12.3, h
SR
, h
RD
and h
SD
are the
channel coefficients which are modeled
as zero-mean, complex Gaussian random
variables with variances and
, and P
R
and P
S
denote the power
transmitted by the relay and the source,
respectively.
Figure 12.3. The three-node
relay channel
2
RD
2
SR
,o o
2
SD
. o
Cooperative relaying
Wireless cooperation involve two phases.
The coordination phase - the strategy is to decide on the best source
node broadcasting method to adopt for its signals being sent to both
destination and relay.
The cooperation phase - the decisions to make involves further
processes on the overhead signals and the method of forwarding them
to the destination.
The terminology full-duplex is used for the relay nodes to transmit
and receive simultaneously [7] compared to the half-duplex setup
[10], [11] where relays cannot transmit and receive simultaneously
in the same band, i.e., relays cannot use the same frequency band
in Frequency division multiple access (FDMA) and orthogonal
frequencies (OFDMA) or the same time-slots in Time division
multiple access (TDMA) systems.
8
Cooperative strategies
Cooperative diversity can be performed based on different relaying
strategies such as:
Amplify and Forward (AF)
allows the relay node to amplify the received noisy signal from the
source node and then forward it to the destination,
simplest relaying strategies with low implementation cost,
provides a better performance when the relay is located half the way
between the source and destination [14].
Decode and Forward (DF)
allows the relay node to decode the received noisy signal from the
source node, re-encode it and forward it to the destination,
DF outperforms AF when the source-relay channel ensures error-free
detection of the received signal at the relay [10], [11],
receiver need CSI between source and relay for optimum decoding.
9
High SNR regime
Zheng and Tse [25] show that:
the outage capacity for the system is equivalent to the diversity-
multiplexing tradeoffs (DMT) in the high SNR. Therefore they propose
to use DMT as a performance measure for various MIMO schemes.
DMT demonstrate that although it is not possible to achieve full diversity
and full multiplexing gains simultaneously but it is possible to use part of
available antennas to increase the data rate and then use the remaining
antennas to increase the error reliability which indicates tradeoffs
between these two gains,
For a transmission scheme to achieve a mutual multiplexing gain r and
diversity gain d at high SNR, it should be able to send data at rate R
(SNR) and an average error probability P
e
(SNR), where both are
functions of channel SNR and satisfy:


10
( )
r
R
>

SNR log
SNR
lim
SNR
( )
d
P
e
s

SNR log
SNR log
lim
SNR
(12.10)
(12.11)
High SNR regime
The optimal tradeoff curve for any scheme operating in Rayleigh fading
channels, and with fading block length l exceeding the total number of transmit
antennas M
T
and the receive antennas M
R
(l M
T
+ M
R
-1), becomes a
piecewise linear function connecting the points (r, d
optimal
(r)). d
optimal
(r) is
defined as the best achievable diversity gain at data rate r, where:





Figure 12.4 shows MIMO scheme
with M
T
= M
R
= 6.
11
(12.12)
(12.13)
( )( )
R T optimal
R T optimal
0
, r r ) (
M M d
M M r d
s s
=
{ }
R T
, min 0
,
) SNR log(
M M r
R
r
s s
=
Figure 12.4. The optimal diversity multiplexing tradeoffs curve d
optimal
(r)
for the general number of m = M
T
, n = M
R
and M
R
(l M
T
+ M
R
-1).
three-node relay channel [25] (IEEE copyright line 2010).

Low SNR regime
Assumptions used during the calculation of the upper bound for high SNR
channel conditions (e.g., DMT curves) are not valid for applications such as
wireless sensor networks that operate within limited bandwidth and energy
resources, hence,
cooperation strategies that can ensure efficient energy transfer, through fading
networks under low SNR channel condition, are needed,
Avestimehr and Tse [19] show that in a simple point to point network,
operating at low SNR and low outage probability the loss in the achieved
capacity given as a ratio between the outage capacity and the AWGN capacity
C
AWGN
, is significantly higher than that observed at high SNR with , as shown
in Figure 12.8,
On solution to improve the outage capacity at low SNR is to employ diversity.
Figure 12.9 shows that the outage capacity (C
I
), of a network operating in low
SNR channels, can be reduced by a factor of , where L is diversity order, and
this improvement is higher at low outage probability.
12
Low SNR regime
13
Figure 12.9. Improvement in the
outage capacity (C
I
) as a function
of the diversity order L under
Rayleigh fading for different
outage probability = 1, 5, 8, and
10%.
Figure 12.8. The loss in the outage
capacity in the fading channel to
AWGN capacity under Rayleigh fading,
for = 0.1 and = 0.01
[19] (IEEE copyright line 2010).

c
C
c
c
Cooperation via Network Coding
14
Cooperation via Network Coding
Network coding protocols
o Packet level NC,
o Physical Layer NC (PNC),
o Analog NC (ANC)
Cooperative network coding (CNC)
o Network spread coding (NSC)
o Network coding and diversity (NCD)


Packet level NC
Existing packet-level NC designs offer limited features for the
current classic signal processing capabilities due to:
increased complexity and cost of encoding and decoding nodes,
which present a major obstacle for NC applications such as wireless
sensor networks, where, NC is probably needed most to reduce the
number of transmissions and preserve the scarce energy supply,
wireless networks are broadcast channels in nature, where, a signal
transmitted by one node may reach several other nodes at the same
time, and a destination node may receive signals from multiple nodes
simultaneously, that can result in excessive interference and therefore
reduction in overall network throughput,
15
Physical layer NC (PNC) and
Analog NC (ANC)
Physical NC (PNC) can make use of additive nature of the wireless
channels to seek higher capacity than the packet-level NC [2],
however, strict synchronization conditions of the PNC limits its use
in practical wireless networks, where signals suffer from variable
delays.
ANC from Katti et al. [3], exploits signal interference at the
intermediate nodes to increase throughput, whilst relaxing the
condition of synchronization between the mixed signals, however,
ANC scheme outperforms packet-level NC, when a high SNR is
assumed with no fading in the communication channel, and the
mixed signals have similar power levels. Otherwise, severe
degradation in performance can be experienced.
16
17
Physical-Layer Wireless Network Coding
Exploit broadcast nature of the wireless link to reduce the
required bandwidth
Information from each user are combined on the signal level
Problem: low noise and no fading assumed and power levels
of the two signals must be the same
User 1
User 2
A
B
A+B
B
A
Cooperative Network coding (CNC)
Xiao et al. study the complex case of combining cooperative
strategies with NC and error control codes [6] introducing the
concept of cooperative network coding (CNC) as a way to address
inefficient resource usage of a network.
In their proposed scheme the transmitting nodes perform some
algebraic superposition of locally generated information prior to that
of the partners encoded using a convolutional code generator matrix.
This system outperforms classic cooperative strategies for its time
sharing and simple message superposition, e.g., using simple XOR
operation for locally generated and relayed bits. With the channel
coding involved, however, this scheme comes at the expense of
increased complexity.
18
Cooperative Network coding
Yu, et al. combine NC with cooperative communication to reduce
overall inter-user interference [5].
This improvement is achieved by increasing the diversity gain in
multiuser fading channels compared to the traditional time-sharing
relaying strategies. For this they consider a scheme that each relay
transmits codewords that contain three parts:
Its own message
Parts of its partners previously transmitted codeword (parity bits) or
codewords
Parity bits generated from joint encoding of the two parts.
19
Network Spread coding (NSC)
In [4], authors proposed the use of the network spread coding (NSC)
scheme as a novel PNC based on spread spectrum using the mutually
orthogonal complete complementary (CC) sequences [47].
NSC allows transmitted signals from different nodes to:
mix in the shared channel virtual network coding,
combats the effects of interference and noise spread spectrum
NSC uses the linearly independent CC sequences to generate local
and global encoding vectors [38] that maintain their orthogonality
over asynchronous communications, high interference, and adverse
channel conditions,
Similar to ANC [3], signal-mixing occurs within a channel at the
physical layer. However, in contrast to ANC and PNC [2], the
proposed NSC scheme can operate at different SNR levels and under
high level of interference caused by the de-synchronization mixing
of signals in mobile fading channels.
20
21
Spread Spectrum
Expand the bandwidth to enable multiple
access and reduce interference, fading and
noise effects
User 1
User 2
A
B
B
A
B
A
NSC- F and NSC-DSF
The proposed NSC scheme can use two operation modes, at
intermediate nodes to facilitate various complexity/cost/performance
trade-offs:
the low-complexity forward mode (NSC-F) where each intermediate
node simply forwards the received mixed signal to its destination
without further processing,
the more error-robust despread-spread-forward mode (NSC-DSF)
where each intermediate node despreads the incoming mixed signals to
recover the transmitted signals and then re-spreads and forwards them to
their destination nodes.
The proposed NSC system brings together NC and spread spectrum
techniques, exploiting the advantages of both, i.e., bandwidth efficiency
of NC and interference and noise robustness of spread spectrum.
22
Network Spread coding (NSC)
Figure 12.11(a)-(c) shows a single-session multicast wireless network
with 2-source nodes sending signals to 2-destination nodes through
one intermediate node (relay) using the:
the general NC scheme [1], which can be either packet-level NC or PNC,
shown in Figure 12.11(a),
the traditional spreading scheme using CC sequences but without NC,
shown in Figure 12.11(b) and,
the proposed NSC scheme in Figure 12.11(c).
23
Figure 12.11. The
example of the
standard
Butterfly-like
wireless network:
(a) a general NC
[1]; (b) a
traditional
spreading scheme
without NC; (c)
the single-session
multicast NSC
(NC + CC).
Network coding and diversity (NCD)
Ding et al. [49] studied the extension of the PNC to the multipath
fading channels, in which pass-loss and phase-shift hinders the PNC
usage, and proposed combining network coding and diversity (NCD),
shown in Figure 12.13.





The NCD scheme is performed in two steps:
In the 1
st
step, the nodes broadcast their messages, with no relaying at this
stage, then the NCD requests the help of the upper medium access layer
(MAC) to sort the available relays according to their local channels
qualities. This ensures better PNC performance.
In the 2
nd
step, then chosen best relay employs AF strategy and broadcasts
the mixture. [49] provides detailed simulation results comparing the outage
capacity and ergodic capacity of the proposed NCD, conventional PNC
scheme, and the direct transmission. 24
Figure 12.13. A diagram of the proposed
NCD transmission strategy [49] (IEEE
copyright line 2010).
Network coding and diversity (NCD)
Figure 12.14 (a) and (b) shows the outage capacity of the three
transmission schemes at different SNR (10 and 20dB), where the
proposed NCD achieves larger outage and ergodic capacities than
those of the conventional PNC in the high SNR channel condition.
25
Figure 12.11. Mutual information complementary cumulative distribution functions for (a)
SNR=10dB, and (b) SNR = 20dB. The distance of the two sources is 2m. Solid line represents the
results obtained by using the Monte-Carlo simulations, and the dotted line represents the results
calculated by using the proposed analytical formulations [49] (IEEE copyright line2008).
(a) SNR = 10dB (b) SNR = 20dB